Using Force Sensors and Neural Models to Encode Tactile Stimuli as Spike-based Responses.

Abstract

Tactile sensors will augment the next generation of prosthetic limbs. However, currently available sensors do not produce biologically-compatible output. This work seeks to illustrate that a force sensor combined with a bi-phasic, neural spiking algorithm, or spiking-sensor, can produce spiking patterns similar to that of the slowly adapting type I (SAI) mechanoreceptor. Experiments were conducted where first spike latency and inter-spike interval, in response to a rapidly delivered (100 ms) sustained displacement (1.1, 1.3, 1.5 mm for 5 s), were compared between the spiking-sensor and SAI recording. The results indicated that the predicted spike times were similar, in magnitude and increasing linear trend, to those observed with the SAI. Over the three displacements, average dynamic ISIs were 7.3, 4.2, 3.8 ms for the spiking-sensor and 6.2, 6.9, 4.1 ms for the SAI, while average static ISIs were 69.0, 45.2, 35.1 ms and 159.9, 69.6, 38.8 ms. The predicted first spike latencies (74.3, 73.9, 96.3 ms) lagged in comparison to those observed for the SAI (26.8, 31.7, 28.8 ms), which may be due to both the different applied force ramp-ups and the SAI's exquisite dynamic sensitivity range and rapid response time.